I’ve been doing a lot of AI Workshops lately with our clients. In a recent one, we were reviewing a scoring process for incoming opportunities. It involved data from a few disparate systems.
At first glance, it seemed like a great candidate for AI. It involved input from several different sources, and we were there to talk about AI, after all. We figured, “Maybe AI could help us make this more intelligent.”
But as we thought about it, the opposite became clear. What we really needed was deterministic outcomes: same inputs, same outputs, every time. We needed to be able to explain exactly how a score was produced. And trust that the logic wouldn’t drift or surprise us later.
If we had used AI, we would’ve introduced variability. Even with prompt engineering or model constraints, generative AI isn’t built for strict consistency. That’s not a bug. That’s what makes AI interesting. But determinism isn’t what it was designed for.
So we backed off.
Not because AI isn’t useful. But because it wasn’t the right tool for the problem.
The Temptation to Overuse AI
This wasn’t a one-off. I’ve noticed a growing pattern across teams and companies: Everyone feels like they should be doing something with AI. And when you feel like you should, you start looking for places to squeeze it in, even where it doesn’t belong.
That’s how you end up using a hammer on a screw. Sure, you might get it to stick, but it’s going to cause a mess.
When AI Isn’t the Answer
AI is great at solving fuzzy problems. It struggles with deterministic outcomes.
Here’s how I think about the difference:
| If your process involves… | You probably need… |
|---|---|
| Rules you can clearly define | Code / logic / decision trees |
| Consistency every time | Deterministic systems |
| Unstructured input (text, images, etc.) | AI / ML |
| Human gut-checks or judgment | AI as a tool, not a decider |
| Clear audit trails or compliance | Deterministic outputs |
Don’t Skip What Works
There’s nothing wrong with AI curiosity. We should be experimenting. But we shouldn’t throw out tried and true methods just because AI is the new kid on the block. Sometimes the right solution is an if statement, not a model.
Final Thought
Using AI well starts with understanding the problem, not chasing the technology. Be clear about the outcome you need. Then pick the tool that helps you get there with the least complexity.
Because not every problem is a nail.